Presaige analyzes images and videos pixel by pixel and predicts how likely they are to generate engagement before they go live. Co-founder David Gioiella demoed the platform at NAB 2026, showing how brands like NBC and Lyft are using confidence scores to pick thumbnails, order carousels, and iterate on generative AI assets.
Key takeaways:
10 million+ image training set correlates pixel-level data with real engagement metrics
Platform-agnostic scoring: no need to select a demographic or distribution platform
Thumbnail selector picks the best frame from a video automatically
Carousel ordering ranks images by predicted engagement for optimal sequence
NBC Olympics and Lyft among the clients using Presaige to optimize deployed creative
Pixel Analysis, Not Content Analysis
Presaige's algorithm does not evaluate what an image depicts. It analyzes pixel-level patterns and correlates them with engagement data from a training library of over 10 million images and videos. The scores reflect how similar visual patterns performed when deployed, independent of subject matter, platform, or target audience. Gioiella was explicit: the system does not care about the concept, aesthetics, or golden ratio. It compares pixels against the dataset and returns a confidence score.
This approach means a user does not need to specify whether the content is for TikTok, Instagram, or broadcast. The algorithm normalizes for platform differences and demographic variation, arguing that a strong image performs well regardless of context.
Thumbnail Selection From Video
The thumbnail selector tool scans every frame of a video and surfaces one to four candidates ranked by predicted engagement. Science YouTuber Jack Gordon, who has been testing the tool, puts his thumbnail options into Presaige and deploys the one with the highest score instead of A/B testing on YouTube. For longer videos, creators focus on scoring the first six seconds to confirm the hook shot rates high, since Presaige limits video analysis to 90 seconds due to processing cost.
Scoring Generative AI Assets
For AI-generated imagery, Presaige offers an apples-to-apples comparison workflow. Gioiella showed a Lyft project where three text-to-image generations in Nano Banana Pro produced slightly different results. Instead of subjective selection, the team deployed the image with the highest Presaige score into video generation, scored three video outputs, and moved the top-scoring one into production. Each result comes with a readiness score on a 1-to-10 scale and specific improvement suggestions for users who want to push the score higher.
From Carousels to Broadcast Promos
NBC used Presaige to optimize a figure skating promo for the Olympics that reached nearly 20 million views on social platforms. Gioiella iterated on the edit using Presaige scores until the piece reached a level he was confident deploying. For carousel posts, the platform scores each image and recommends the optimal sequence. Presaige is also building integrations with media management platforms like Simeon and Freepik, where the scoring would run inside existing review and approval workflows via API.
Presaige offers a free tier with image scoring, a premium tier that unlocks video analysis, and an enterprise tier with custom pricing and full API access.
